app.py

app.py

@@ -7,18 +7,8 @@ from transformers import Wav2Vec2Processor, AutoModelForCTC
 
 nltk.download("punkt")
 
-wav2vec2_processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-base-960h") 
-wav2vec2_model = AutoModelForCTC.from_pretrained("facebook/wav2vec2-base-960h")
-hubert_processor = Wav2Vec2Processor.from_pretrained("facebook/hubert-large-ls960-ft") 
-hubert_model = AutoModelForCTC.from_pretrained("facebook/hubert-large-ls960-ft")
-
 def return_processor_and_model(model_name):
-    if model_name == "facebook/wav2vec2-base-960h":
-        return wav2vec2_processor, wav2vec2_model
-    elif model_name == "facebook/hubert-large-ls960-ft":
-        return hubert_processor, hubert_model
-    else:
-        return None
+    return Wav2Vec2Processor.from_pretrained(model_name), AutoModelForCTC.from_pretrained(model_name)
 
 def load_and_fix_data(input_file):  
   #read the file
@@ -72,6 +62,6 @@ gr.Interface(return_all_predictions,
              inputs = [gr.inputs.Audio(source="microphone", type="filepath", label="Record/ Drop audio"), gr.inputs.Dropdown(["facebook/wav2vec2-base-960h", "facebook/hubert-large-ls960-ft"], label="Model Name")],
              outputs = [gr.outputs.Textbox(label="Beam CTC decoding"), gr.outputs.Textbox(label="Greedy decoding")],
              title="ASR using Wav2Vec2/ Hubert & pyctcdecode",
-             description = "Comparing greedy decoder with beam search CTC decoder (https://distill.pub/2017/ctc/), record/ drop your audio!",
+             description = "Comparing greedy decoder with beam search CTC decoder, record/ drop your audio!",
              layout = "horizontal",
              examples = [["test1.wav", "facebook/wav2vec2-base-960h"], ["test2.wav", "facebook/hubert-large-ls960-ft"]], theme="huggingface").launch()